Deploy and Host Hermes Agent (Slim) on Railway

Hermes is a self-improving AI agent by Nous Research — it connects to your messaging channels, learns from every interaction, creates its own skills, and gets more capable over time. This Slim template runs it headless (CLI + always-on messaging gateway, no web UI) for the smallest possible footprint — ~112 MB RAM.

About Hosting Hermes Agent (Slim)

Hosting the Slim variant means running an always-on hermes gateway that answers Telegram, Discord, and Slack around the clock, while persisting skills, memories, sessions, and auth on a Railway volume (/data) so they survive redeploys. It ships no dashboard, no Node, and no headless browser — just the agent and its messaging gateway, with glibc tuned (MALLOC_ARENA_MAX=2) so the always-on engine idles around ~112 MB. Because Railway bills by GB-hour of memory, a leaner agent means a lower monthly bill for the same 24/7 self-improving agent. You can drive it from your terminal over railway ssh + tmux and hermes chat.

Why Deploy Hermes Agent (Slim) on Railway?

Railway charges for the RAM you use, and this variant is built to use as little as possible:

	this template (slim)	typical Hermes template	browser-bundled template
Steady RAM	~112 MB	~300–500 MB	1–4 GB
Web dashboard	none (CLI)	yes	yes
Node / Playwright / Chromium	none	sometimes	yes

Lowest RAM, lowest bill — the leanest way to keep a Hermes agent online 24/7.
Self-improving — agent-curated memory, autonomous skill creation, cron jobs.
Multi-channel — Telegram, Discord, Slack, and more.
Your keys, your data — bring an API key or a ChatGPT/Codex subscription; nothing is bundled and no credentials ship with the template.

Common Use Cases

A 24/7 personal assistant on Telegram or Discord that remembers context and improves over time.
A self-improving ops/support bot that builds its own skills from real tasks.
A cost-sensitive always-on agent where minimizing Railway memory spend matters.

Dependencies for Hermes Agent (Slim) Hosting

An LLM provider — set an API-key variable (OPENAI_API_KEY, OPENROUTER_API_KEY, ANTHROPIC_API_KEY, …), or authenticate a ChatGPT/Codex subscription over railway ssh:
```
railway ssh -s hermes -- hermes auth add openai-codex --type oauth --no-browser
```
(optional) a messaging bot token — TELEGRAM_BOT_TOKEN, DISCORD_BOT_TOKEN, or SLACK_BOT_TOKEN — to bring an always-on channel online.

Deployment Dependencies

Hermes Agent (upstream source) — built from @main, pinnable.
A Railway volume mounted at /data (attached automatically) for persistent skills, memories, sessions, and auth.

Variables

Variable	Required	Purpose
`OPENAI_API_KEY` / `OPENROUTER_API_KEY` / `ANTHROPIC_API_KEY`	optional*	LLM provider key (*or use the Codex device-flow above)
`TELEGRAM_BOT_TOKEN` / `DISCORD_BOT_TOKEN` / `SLACK_BOT_TOKEN`	optional	enable an always-on messaging channel
`HERMES_KEEPALIVE`	optional	`auto` (default), `gateway`, or `idle`

Low-memory tuning (MALLOC_ARENA_MAX=2) ships baked into the image — nothing to set.

Want a point-and-click web dashboard instead? A Full variant — the first-party Nous dashboard, built from source — is also available on this marketplace.

Hermes Agent is open source by Nous Research. This template bundles no third-party UI — only Hermes-team artifacts.